On 11/10/25 5:37 PM, Akhil P Oommen wrote:
> A8x GMU configuration are very similar to A7x. Unfortunately, there are
> minor shuffling in the register offsets in the GMU CX register region.
> Apart from that, there is a new HFI message support to pass table like
> data. This patch adds support for perf table using this new HFI
> message.
>
> Apart from that, there is a minor rework in a6xx_gmu_rpmh_arc_votes_init()
> to simplify handling of MxG to MxA fallback along with the additional
> calculations for the new dependency vote.
>
> Signed-off-by: Akhil P Oommen <[email protected]>
> ---
[...]
> /* Make sure there are no outstanding RPMh votes */
> @@ -990,7 +1022,8 @@ static void a6xx_gmu_rpmh_off(struct a6xx_gmu *gmu)
> gmu_poll_timeout_rscc(gmu, REG_A6XX_RSCC_TCS3_DRV0_STATUS + seqmem_off,
> val, (val & 1), 100, 1000);
>
> - if (!adreno_is_a740_family(adreno_gpu))
> +
> + if (!adreno_is_a740_family(adreno_gpu) && !adreno_is_a8xx(adreno_gpu))
stray double \n above
[...]
> /* Check to see if we are doing a cold or warm boot */
> - if (adreno_is_a7xx(adreno_gpu)) {
> + if (adreno_is_a7xx(adreno_gpu) || adreno_is_a8xx(adreno_gpu)) {
> status = a6xx_llc_read(a6xx_gpu, REG_A7XX_CX_MISC_TCM_RET_CNTL)
> == 1 ?
You skipped writing to this register above - can we count on it
always being in the right state?
Do we need to map VRB, or is that a topic for the future?
Konrad
